Neighbor Page and One Settings

Submitted by estefan31 on Tue, 08/30/2022 - 01:19

My main question here is: should the simulation results be dependent on the values chosen for the neighbor page and one settings?

My neighbor settings are set as follows with the skin set to the diameter of the biggest particle in my simulation, and my particles are uniform or in some cases have a small range of sizes:

neighbor ${maximum_particle_diameter} bin
neigh_modify delay 0 every 1 check yes page 100000 one 2000

With the settings above, I noticed no issues. But if I set "one 4000" or higher, I get results that are very different, as seen in the attached file.

I am simulating the direct shear test with about 10,000 to 50,000 particles. If I set "one 2000" then all my stress-strain results are about the same at very slow quasi-static velocities. However, if I set "one 4000" "one 6000" or "one 8000", my peak stresses in the stress-strain results become heavily dependent on my shear velocity, all other parameters kept the same. I'm finding that literally the only difference in my simulation results is if I set "neigh_modify one 2000" or "neigh_modify one 8000" and these different "one" settings cause shear behavior that is relatively independent of shear velocity (assuming slow shear velocities) and shear behavior that is highly dependent on shear velocity.

Can someone tell me what might be going on here, and how can I increase my one settings for memory purposes without dramatically changing my simulation results? The memory issues arise when I start using a wider range of particle sizes in my project.

estefan31 | Tue, 08/30/2022 - 01:24

As for the boundary conditions, I'm using mesh walls for the boundaries of the shear box, including a servo-wall for the upper plate to maintain a constant vertical stress.

estefan31 | Mon, 09/12/2022 - 20:08

There's a bug in neighbor.cpp. I recommend just changing the default one and page settings and recompiling the code if you're running into memory issues.